Spelling suggestions: "subject:"topic amodeling"" "subject:"topic emodeling""
31 |
Large-scale Exploratory Text VisualisationAxelsson, Wilma, Engström, Nellie January 2023 (has links)
The amount of available text data has increased rapidly in the latest years, making it difficult for an everyday user to find relevant information. To solve this, NLP and visualisation methods have been developed for extracting valuable information from text and presenting it to the user. The aim of this project is to implement a proof-of-concept visualisation prototype for exploring a large amount of Swedish news articles with related metadata and investigate the temporal and relational aspects of the data. The project was divided into three major parts. In the first part, sketches of the visualisation were designed and evaluated through user tests. The second part consisted of designing and implementing a NLP pipeline, using BERTopic, where both Dynamic Topic Modeling (DTM) and Hierarchical Topic Modeling (HTM) were used. Some parameters of the pipeline were evaluated using evaluation metrics and through visual inspection, for instance a Swedish sentence transformer. The final part consisted of implementing and evaluating the visualisation prototype. The project resulted in a web-based visualisation, presenting the NLP results, with two different views: a top 10 topics view and a hierarchical view containing all topics. The prototype has various features, e.g., clicking and hovering for details-on-demand and options for changing and altering the view. The prototype was then evaluated through an internal case study and user tests. For the user tests, there were two groups of participants: people working in the journalism field and people working closely to the NLP field. Both groups thought there was more value in viewing the top 10 topics view than the hierarchical view. Furthermore, the quality of the top 10 topics view was considered higher overall compared to the hierarchical view. In the end, the result of this project is a proof-of-concept visualisation prototype presenting topics of Swedish news articles, over time and in relation to each other. A few possible improvement possibilities include improving the hierarchical relations between the topics and the run time of the topic model and prototype. Also, the prototype may be further improved with additional features, e.g., real-time data, a map, the full text of the articles and a search function. / <p>Examensarbetet är utfört vid Institutionen för teknik och naturvetenskap (ITN) vid Tekniska fakulteten, Linköpings universitet</p>
|
32 |
Digital Maturity in the Public Sector and Citizens’ Sentiment Towards Authorities : A study within the initiative Academy of Lifelong Learning, in partnership with RISE and GoogleCramner, Isabella January 2021 (has links)
This study was conducted in partnership with RISE and Google, within the initiative “Academy of Lifelong Learning”, aiming to propel the digital transformation in the Swedish public sector. The study investigated the digital maturity of 18 authorities in terms of maturity level (early, developing maturing), and within the driving areas (1) Citizen Centricity, (2) Leadership, (3) Digital Toolbox and (4) Security and Sustainability. Further, it explored how citizens’ sentiment towards public authorities relates to the organizations’ digital maturity scores. The results of a digital maturity survey showed that 16 of the 18 contributing organizations were developing, whereas two scored just enough to be classified as maturing. The organizations performed best within Security and Sustainability, and the worst within the category Digital Toolbox—where the biggest competence gaps were also identified. To unlock citizens’ sentiment towards the authorities, sentiment analysis was conducted on Facebook data. In a correlation analysis, a significant negative relationship was surprisingly found between (i) maturity score and (ii) sentiment score, as well as between (i) maturity score and (ii) positive comments. Presumably, this can be explained by citizens interacting the most with the more mature organizations and thus expressing their dissatisfaction more. However, more analysis is needed to draw conclusions. / Studien genomfördes i samarbete med RISE och Google inom initiativet ”Akademin för livslångt lärande” (Academy of Lifelong Learning), som syftar till att driva på den digitala transformationen i den svenska offentliga sektorn. Studien undersökte 18 myndigheters digitala mognad med fokus på mognadsnivå (early, developing maturing), och inom de drivande områdena (1) medborgarperspektivet, (2) ledarskap, (3) digitala verktygslådan och (4) säkerhet och hållbarhet. Vidare undersöktes medborgarnas attityder gentemot offentliga myndigheter i relation till organisationernas digitala mognad. Resultatet från mognadsundersökningen visade att 16 av de 18 medverkande organisationerna var developing, medan två organisationer precis kunde klassificeras som mature. Organisationerna presterade bäst inom säkerhet och hållbarhet och sämst inom kategorin digitala verktygslådan—där de största kompetensbristerna även identifierades. För att utvärdera medborgarnas attityder gentemot myndigheterna genomfördes en sentimentanalys baserat på data från Facebook. I en korrelationsanalys hittades överraskande nog en signifikant negativt samband mellan (i) digital mognad och (ii) sentimentpoäng, samt mellan (i) digital mognad och (ii) positiva kommentarer. Detta kan antagligen förklaras med att medborgarna interagerar mer med de mest mogna organisationerna och därmed är mer benägna att utrycka sitt missnöje gentemot dem. Ytterligare analys behövs dock för att kunna dra sådana slutsatser och förklara resultatet.
|
33 |
Nyhetsmedierna om Trumps valkampanj : En diskursanalys av 3652 artiklar genom topic modeling med MALLET / News media on the Trump campaign : A discourse analysis of 3652 news articles using topic modeling through MALLETÅkerlund, Mathilda January 2017 (has links)
The aim of this study was to examine how American news media covered Donald Trump's presidential campaign in the election of 2016, as well as discussing the possible consequences of such reporting on the election results. Using mixed methods, 3652 digital news articles were studied by discourse analysis and topic modeling through MALLET. The study found that a substantial number of articles were dedicated to such non-political news reporting as scandals, portraying an image of Trump as someone who can get away with doing whatever he wants. Furthermore, the results of the study found that media helped to convey Trump’s views of minorities, doing so in particularly by citing him. The media also relied largely on polls. Comparison of the candidates through these polls enhanced the image of the election campaign as nothing more than a horse race, as well as turning up Trumps entertainment value. As the campaign continued, the reporting got more aggressive towards Trump. At the same time there was an element of wanting to balance the critical articles about him by simultaneously writing negatively about other candidates. The study concludes that all of the non-political new stories might have directed focus away from the important policy issues, leading to people voting for candidates without the proper insight into their politics.
|
34 |
LDA based approach for predicting friendship links in live journal social networkParimi, Rohit January 1900 (has links)
Master of Science / Department of Computing and Information Sciences / Doina Caragea / The idea of socializing with other people of different backgrounds and cultures excites the web surfers. Today, there are hundreds of Social Networking sites on the web with millions of users connected with relationships such as "friend", "follow", "fan", forming a huge graph
structure. The amount of data associated with the users in these Social Networking sites has resulted in opportunities for interesting data mining problems including friendship link and interest predictions, tag recommendations among others. In this work, we consider the friendship link prediction problem and study a topic modeling approach to this problem.
Topic models are among the most effective approaches to latent topic analysis and mining of text data. In particular, Probabilistic Topic models are based upon the idea that documents can be seen as mixtures of topics and topics can be seen as mixtures of words. Latent Dirichlet Allocation (LDA) is one such probabilistic model which is generative in nature
and is used for collections of discrete data such as text corpora. For our link prediction
problem, users in the dataset are treated as "documents" and their interests as the document contents. The topic probabilities obtained by modeling users and interests using LDA provide an explicit representation for each user. User pairs are treated as examples and are represented using a feature vector constructed from the topic probabilities obtained with LDA. This vector will only capture information contained in the interests expressed by the
users. Another important source of information that is relevant to the link prediction task is given by the graph structure of the social network. Our assumption is that a user "A"
might be a friend of user "B" if a) users "A" and "B" have common or similar interests
b) users "A" and "B" have some common friends. While capturing similarity between
interests is taken care by the topic modeling technique, we use the graph structure to find common friends. In the past, the graph structure underlying the network has proven to be a
trustworthy source of information for predicting friendship links. We present a comparison of predictions from feature sets constructed using topic probabilities and the link graph
separately, with a feature set constructed using both topic probabilities and link graph.
|
35 |
Techniques d'identification d'entités nommées et de classification non-supervisée pour des requêtes de recherche web à l'aide d'informations contenues dans les pages web visitéesGoulet, Sylvain January 2014 (has links)
Le web est maintenant devenu une importante source d’information et de divertissement pour un grand nombre de personnes et les techniques pour accéder au contenu désiré ne cessent d’évoluer. Par exemple, en plus de la liste de pages web habituelle, certains moteurs de recherche présentent maintenant directement, lorsque possible, l’information recherchée par l’usager. Dans ce contexte, l’étude des requêtes soumises à ce type de moteur de recherche devient un outil pouvant aider à perfectionner ce genre de système et ainsi améliorer l’expérience d’utilisation de ses usagers. Dans cette optique, le présent document présentera certaines techniques qui ont été développées pour faire l’étude des requêtes de recherche web soumises à un moteur de recherche. En particulier, le travail présenté ici s’intéresse à deux problèmes distincts. Le premier porte sur la classification non-supervisée d’un ensemble de requêtes de recherche web dans le but de parvenir à regrouper ensemble les requêtes traitant d’un même sujet. Le deuxième problème porte quant à lui sur la détection non-supervisée des entités nommées contenues dans un ensemble de requêtes qui ont été soumises à un moteur de recherche. Les deux techniques proposées utilisent l’information supplémentaire apportée par la connaissance des pages web qui ont été visitées par les utilisateurs ayant émis les requêtes étudiées.
|
36 |
Nonparametric Discovery of Human Behavior Patterns from Multimodal DataSun, Feng-Tso 01 May 2014 (has links)
Recent advances in sensor technologies and the growing interest in context- aware applications, such as targeted advertising and location-based services, have led to a demand for understanding human behavior patterns from sensor data. People engage in routine behaviors. Automatic routine discovery goes beyond low-level activity recognition such as sitting or standing and analyzes human behaviors at a higher level (e.g., commuting to work). The goal of the research presented in this thesis is to automatically discover high-level semantic human routines from low-level sensor streams. One recent line of research is to mine human routines from sensor data using parametric topic models. The main shortcoming of parametric models is that they assume a fixed, pre-specified parameter regardless of the data. Choosing an appropriate parameter usually requires an inefficient trial-and-error model selection process. Furthermore, it is even more difficult to find optimal parameter values in advance for personalized applications. The research presented in this thesis offers a novel nonparametric framework for human routine discovery that can infer high-level routines without knowing the number of latent low-level activities beforehand. More specifically, the frame-work automatically finds the size of the low-level feature vocabulary from sensor feature vectors at the vocabulary extraction phase. At the routine discovery phase, the framework further automatically selects the appropriate number of latent low-level activities and discovers latent routines. Moreover, we propose a new generative graphical model to incorporate multimodal sensor streams for the human activity discovery task. The hypothesis and approaches presented in this thesis are evaluated on public datasets in two routine domains: two daily-activity datasets and a transportation mode dataset. Experimental results show that our nonparametric framework can automatically learn the appropriate model parameters from multimodal sensor data without any form of manual model selection procedure and can outperform traditional parametric approaches for human routine discovery tasks.
|
37 |
Topic Analysis of Hidden Trends in Patented Features Using Nonnegative Matrix FactorizationLin, Yicong 01 January 2016 (has links)
Intellectual property has gained more attention in recent decades because innovations have become one of the most important resources. This paper implements a probabilistic topic model using nonnegative matrix factorization (NMF) to discover some of the key elements in computer patent, as the industry grew from 1990 to 2009. This paper proposes a new “shrinking model” based on NMF and also performs a close examination of some variations of the base model. Note that rather than studying the strategy to pick the optimized number of topics (“rank”), this paper is particularly interested in which factorization (including different kinds of initiation) methods are able to construct “topics” with the best quality given the predetermined rank. Performing NMF to the description text of patent features, we observe key topics emerge such as “platform” and “display” with strong presence across all years but we also see other short-lived significant topics such as “power” and “heat” which signify the saturation of the industry.
|
38 |
Holy day effects on language: How religious geography, individual affiliation and day of the week relate to sentiment and topics on TwitterKramer, Stephanie 10 April 2018 (has links)
Religious belief and attendance predict improved well-being at the individual level. Paradoxically, geographic locations with high rates of religious belief and attendance are often those with the differentially high rates of societal instability and suffering. Many of the consequences of religiosity are context-based and vary across time, and holy days are naturally-occurring religious cues that have been shown to influence religiously-relevant attitudes and behaviors. I investigated the degree to which personal religiosity and religious geography (i.e. religious demographics with other location variables) individually and interactively predict well-being across days of the week.
In the first study, American Christians demonstrated greater well-being by expressing more positive sentiment in Twitter posts, while American Muslims displayed less well-being. Sundays were generally the most positive day, but American Muslims communicated more happiness on Fridays (the Muslim holy day). In the second study, Christianity did not predict increased well-being in the posts of college students. In the third study, global survey data with measures of religiosity and well-being indicated that the well-being consequences of religious affiliation depend on the religious group and location, and that people tend to be especially positive on their group’s holy day. Study four explored the latent topical content of Twitter posts. Across studies, religious minority status appeared to have a deleterious effect on well-being.
|
39 |
A probabilistic and incremental model for online classification of documents : DV-INBCRodrigues, Thiago Fredes January 2016 (has links)
Recentemente, houve um aumento rápido na criação e disponibilidade de repositórios de dados, o que foi percebido nas áreas de Mineração de Dados e Aprendizagem de Máquina. Este fato deve-se principalmente à rápida criação de tais dados em redes sociais. Uma grande parte destes dados é feita de texto, e a informação armazenada neles pode descrever desde perfis de usuários a temas comuns em documentos como política, esportes e ciência, informação bastante útil para várias aplicações. Como muitos destes dados são criados em fluxos, é desejável a criação de algoritmos com capacidade de atuar em grande escala e também de forma on-line, já que tarefas como organização e exploração de grandes coleções de dados seriam beneficiadas por eles. Nesta dissertação um modelo probabilístico, on-line e incremental é apresentado, como um esforço em resolver o problema apresentado. O algoritmo possui o nome DV-INBC e é uma extensão ao algoritmo INBC. As duas principais características do DV-INBC são: a necessidade de apenas uma iteração pelos dados de treino para criar um modelo que os represente; não é necessário saber o vocabulário dos dados a priori. Logo, pouco conhecimento sobre o fluxo de dados é necessário. Para avaliar a performance do algoritmo, são apresentados testes usando datasets populares. / Recently the fields of Data Mining and Machine Learning have seen a rapid increase in the creation and availability of data repositories. This is mainly due to its rapid creation in social networks. Also, a large part of those data is made of text documents. The information stored in such texts can range from a description of a user profile to common textual topics such as politics, sports and science, information very useful for many applications. Besides, since many of this data are created in streams, scalable and on-line algorithms are desired, because tasks like organization and exploration of large document collections would be benefited by them. In this thesis an incremental, on-line and probabilistic model for document classification is presented, as an effort of tackling this problem. The algorithm is called DV-INBC and is an extension to the INBC algorithm. The two main characteristics of DV-INBC are: only a single scan over the data is necessary to create a model of it; the data vocabulary need not to be known a priori. Therefore, little knowledge about the data stream is needed. To assess its performance, tests using well known datasets are presented.
|
40 |
Fast Inference for Interactive Models of TextLund, Jeffrey A 01 September 2015 (has links)
Probabilistic models of text are a useful tool for enabling the analysis of large collections of digital text. For example, Latent Dirichlet Allocation can quickly produce topical summaries of large collections of text documents. Many important uses cases of such models include human interaction during the inference process for these models of text. For example, the Interactive Topic Model extends Latent Dirichlet Allocation to incorporate human expertiese during inference in order to produce topics which are better suited to individual user needs. However, interactive use cases of probabalistic models of text introduce new constraints on inference - the inference procedure must not only be accurate, but also fast enough to facilitate human interaction. If the inference is too slow, then the human interaction will be harmed, and the interactive aspect of the probalistic model will be less useful. Unfortunately, the most popular inference algorithms in use today either require strong approximations which can degrade the quality of some models, or require time-consuming sampling. We explore the use of Iterated Conditional Modes, an algorithm which is able to obtain locally optimal maximum a posteriori estimates, as an alternative to popular inference algorithms such as Gibbs sampling or mean field variational inference. Iterated Conditional Modes algorithm is not only fast enough to facilitate human interaction, but can produce better maximum a posteriori estimates than sampling. We demonstrate the superior performance of Iterated Conditional Modes on a wide variety of models. First we use a DP Mixture of Multinomials model applied to the problem of web search result cluster, and show that not only can we outperform previous methods in clustering quality, but we can achieve interactive runtimes when performing inference with Iterated Conditional Modes. We then apply Iterated Conditional Modes to the Interactive Topic Model. Not only is Iterated Conditional Modes much faster than the previous published Gibbs sampler, but we are better able to incorporate human feedback during inference, as measured by accuracy on a classification task using the resultant topic model. Finally, we utilize Iterated Conditional Modes with MomResp, a model used to aggregate multiple noisy crowdsourced data. Compared with Gibbs sampling, Iterated Conditional Modes is better able to recover ground truth labels from simulated noisy annotations, and runs orders of magnitude faster.
|
Page generated in 0.0831 seconds